PaCMAP Benchmark

PaCMAP Benchmark #

Load data #

 Dataset (df) contains 333352 columns (5mC nucleotides/probes) and 3330 rows (samples).

Benchmark PaCMAP #

Show code cell source Hide code cell source

import numpy as np
import pacmap
import warnings
from bokeh.plotting import figure, show
from bokeh.transform import factor_cmap
from bokeh.models import Div, Slider, TabPanel, Tabs, Legend, ColumnDataSource
from bokeh.layouts import layout
from bokeh.io import curdoc, output_notebook

class DataProcessor:
    def __init__(self, df_labels, df_methyl, clinical_trials, sample_types, cols, remove_duplicates=True):
        self.df_labels = df_labels
        self.df_methyl = df_methyl
        self.clinical_trials = clinical_trials
        self.sample_types = sample_types
        self.cols = cols
        self.remove_duplicates = remove_duplicates

    def filter_data(self):
        self.df1 = self.df_labels[self.df_labels['Clinical Trial'].isin(self.clinical_trials)]
        self.df2 = self.df1[self.df1['Sample Type'].isin(self.sample_types)]
        if self.remove_duplicates:
            self.df3 = self.df2[~self.df2['Patient_ID'].duplicated(keep='last')]
        else:
            self.df3 = self.df2
        self.df_methyl_filtered = self.df_methyl[self.df_methyl.index.isin(self.df3.index)].iloc[:, 1:]


    def apply_pacmap(self):

        # Define a filter to ignore the specific UserWarning
        warnings.filterwarnings("ignore", category=UserWarning, message="Warning: random state is set to")

        reducer = pacmap.PaCMAP(n_components=2, n_neighbors=15, MN_ratio=0.4, FP_ratio=16.0, 
                                random_state=42, lr=0.1, num_iters=5000)
        embedding = reducer.fit_transform(self.df_methyl_filtered.to_numpy(dtype='float16'))

        self.df_embedding = pd.DataFrame(embedding, index=self.df_methyl_filtered.index, columns=['PaCMAP 1', 'PaCMAP 2'])

    def join_labels(self):
        self.df = self.df_embedding.join(self.df_labels[self.cols]).reset_index()


class BokehPlotter:
    def __init__(self, df, cols, custom_color_palette, title=None, x_range=None, y_range=None):
        self.df = df
        self.cols = cols
        self.custom_color_palette = custom_color_palette
        self.title = title + ', n=' + str(self.df.shape[0])
        self.x_range = x_range or (-50, 50)
        self.y_range = y_range or (-50, 50)
        self.tabs = None
        self.points = None
        self.slider = None
        self.layout = None

    def create_figure(self):
        return figure(title=self.title,
                      width=1200, height=600, sizing_mode='fixed',
                      x_axis_label='PaCMAP 1', y_axis_label='PaCMAP 2',
                      x_range=self.x_range, y_range=self.y_range,
                      tools="pan,wheel_zoom,reset,save", active_drag="pan",
                      active_scroll="auto",
                      tooltips=[("Karyotype", "@Karyotype"),
                                ("Gene Fusion", "@{Gene Fusion}"),
                                ("Patient ID", "@{Patient_ID}")])

    def create_scatters(self, p, hue):
        df = self.df[~self.df[hue].isna()]  # Filter out rows with NaN values for the hue column
        filtered_dfs = [df[df[hue] == val] for val in df[hue].value_counts().sort_values(ascending=False).index.to_list()]
        
        renderers = []
        items = []
        for i in range(len(filtered_dfs)):
            name = filtered_dfs[i][hue].head(1).values[0]
            color = self.custom_color_palette[i % len(self.custom_color_palette)]
            source = ColumnDataSource(filtered_dfs[i])
            r = p.scatter(x="PaCMAP 1", y="PaCMAP 2", source=source,
                         fill_alpha=0.8, size=5,
                         color=color)
            renderers.append(r)
            items.append((name, [r]))

        return renderers, items

    def create_tabs_and_points(self):
        self.tabs = Tabs(tabs=[TabPanel(child=self.create_figure(), title=title) for title in self.cols[:-3]],
                tabs_location='left')

        self.points = [self.create_scatters(tab.child, hue=col) for tab, col in zip(self.tabs.tabs, self.cols[:-3])]

    def finalize_tabs(self):
        for p, (renderers, items) in zip(self.tabs.tabs, self.points):
            p.child.toolbar.logo = None
            p.child.toolbar_location = 'above'
            legend = Legend(items=items, location='top_left')
            p.child.add_layout(legend, 'right')
            p.child.legend.click_policy = 'hide'
        for i in range(len(self.tabs.tabs)):
            self.tabs.tabs[i].child.legend.title = self.tabs.tabs[i].title
            self.tabs.tabs[i].child.output_backend = "svg"

    def create_slider(self):
        self.slider = Slider(title="Adjust datapoint size", start=0, end=20, step=1, value=self.points[0][0][0].glyph.size)
        for i in range(len(self.points)): 
            for r in self.points[i][0]: 
                self.slider.js_link("value", r.glyph, "size")

    def create_layout(self):
        div = Div(text="""<br>""", width=1000, height=10)
        self.layout = layout([[[div, self.tabs, self.slider]]])

    def plot(self):
        self.create_tabs_and_points()
        self.finalize_tabs()
        self.create_slider()
        self.create_layout()
        show(self.layout)

cols         = ['PaCMAP Output', 'ELN22 Final Diagnosis', 'FAB', 'FLT3 ITD', 'Age group (years)',
                'Complex Karyotype', 'Primary Cytogenetic Code' ,'Batch', 'Sex', 'MRD 1 Status',
                'Leucocyte counts (10⁹/L)', 'Risk Group', 'Race or ethnic group',
                'Clinical Trial', 'Vital Status', 'Sample Type', 'Karyotype', 'Gene Fusion', 'Patient_ID']

custom_color_palette = [
    '#1f77b4',
    '#ff7f0e', 
    '#2ca02c',
    '#d62728',
    '#9467bd', 
    '#7f7f7f',
    '#e377c2',
    '#e7ba52',
    '#bcbd22',
    '#17becf',
    '#393b79',
    '#8c564b',
    '#f7b6d2',
    '#c49c94',
    '#e377c2'] 

# Reference: 
# https://github.com/d3/d3-3.x-api-reference/blob/master/Ordinal-Scales.md#categorical-colors

output_notebook()
# Set the theme for the plot
curdoc().theme = 'light_minimal' # or 'dark_minimal'

Loading BokehJS ...

All AML and ALL samples #

Adult and Peds AML samples #

df_labels['Clinical Trial'].value_counts()

Clinical Trial
NOPHO ALL92-2000            933
AAML0531                    628
AAML1031                    581
Beat AML Consortium         316
TCGA AML                    194
CETLAM SMD-09 (MDS-tAML)    166
French GRAALL 2003–2005     153
TARGET ALL                  141
AAML03P1                     72
Japanese AML05               64
CCG2961                      41
Name: count, dtype: int64

Peds AML samples #

AAML1031, 0531, and 03P1 samples #

Table with all patients #

	ELN22 Final Diagnosis	FAB	FLT3 ITD	Age group (years)	Complex Karyotype	Primary Cytogenetic Code	Batch	Sex	MRD 1 Status	Leucocyte counts (10⁹/L)	Risk Group	Race or ethnic group	Clinical Trial	Vital Status	Sample Type	Karyotype	Gene Fusion
Patient_ID
Loading... (need help?)

Watermark #

Author: Francisco_Marchi@Lamba_Lab_UF

Python implementation: CPython
Python version       : 3.8.16
IPython version      : 8.12.2

numpy  : 1.24.3
pandas : 2.0.2
bokeh  : 3.1.1
pacmap : 0.7.0
itables: 1.5.2

Compiler    : GCC 11.3.0
OS          : Linux
Release     : 5.15.90.1-microsoft-standard-WSL2
Machine     : x86_64
Processor   : x86_64
CPU cores   : 20
Architecture: 64bit

PaCMAP Benchmark

Contents